May the structure information of protein be obtained from the corresponding nucleotide sequence? For this question, a computer program was used to conduct a statistical analysis about the clustering phenomena of amino acids. In this method,20 kinds of amino acid were classified into 2 types according to the number of hydrogen bonds formed by middle base of their correlation codons.It can be seen that amino acid has a rather great possibility of neighboring on another of its class and this assembly has a tendency to forming specific secondary structure. A sequence was designed and its secondary structure was predicted by this prediction software.